Parallel Optimal Deep Echo State Network

Introduction

This repository is a generic implementation and example of "Parallel Optimal Deep Echo State Network", but not the full implementation of the article "LEOPARD: Parallel Optimal Deep Echo State Network Prediction Improves Service Coverage for UAV-Assisted Outdoor Hotspots". Therefore, you can use this repository to train your model or as a benchmark for comparison. The system architecture was presented in this paper [LEOPARD] [Slides].

The implementation of DeepESN is based on lucapedrelli-DeepESN and Xeeshanmalik-deep_ml_esn.

The architecture of parallel computing is based on zblanks-parallel_esn.

There are some limitations to this work. If you have any questions or suggestions, please feel free to contact me. Your suggestions are greatly appreciated.

Citing

Please consider citing our paper if this repository is helpful to you. Bibtex:

@INPROCEEDINGS{9548955,
  author={Peng, Haoran and Tsai, Ang-Hsun and Wang, Li-Chun and Han, Zhu},
  booktitle={IEEE Trans. Cogn. Commun. Netw.}, 
  title={{LEOPARD}: Parallel Optimal Deep Echo State Network Prediction Improves Service Coverage for {UAV}-Assisted Outdoor Hotspots},
  volume={8},
  number={1},
  pages={282--295},
  year={2022},
  month = {Mar.}
}

Requirements

• Python: 3.6.13
• xlwt
• bayesian-optimization
• numpy: <= 1.17
• matplotlib
• pandas

Usage

Descriptions of the data files

• The files "lat" and "lon" are the training data sampled from the GeoLife GPS Trajectory Dataset. The last 100 data samples in each file are used for testing and prediction.
• The files "lat_result.xls" and "lon_result.xls" are the predicted results.
• The files "params_lat" and "params_lon" are the optimized paramaters find by Bayesian Optimization.
• The files "deepESN_loss_lat" and "deepESN_loss_lon" are the recorded loss during the training process.

Descriptions of python files

The deep echo state network is impletemented in 'DeepESN.py'. The parallel computing and Bayesian Optimization are implemented in 'train.py'. The file 'predict.py' is for testing.

Training phase

1. For training model by the data lat.txt

python train.py lat

2. For training model by the data lon.txt

python train.py lon

Testing phase

1. For testing model using the data lat.txt

python predict.py lat

2. For testing model using the data lon.txt

python predict.py lon